Disclaimer: I have not yet checked with Jacy because I cannot figure out a parsable input with all the correct whitespace etc.
I am getting the following ambiguity from my current version of the Matrix:
The sentence is supposed to mean Sandy knows what Kim read or something like that.
Is the ambiguity spurious? Or does it perhaps make sense, in the absence of an (optional) question particle, that this sentence may be interpreted in a couple of ways?
I suspect this ambiguity is legit, for a wh-in-situ language. If there are multiple clauses and neither is marked by anything that says “this clause can’t be a question” or “this clause can’t be a wh question” then the wh word can be discharged at either level.
I defer to Emily for the general answer to your question, but I’ll address the part about Japanese and parsing with Jacy (we should really hook up a Japanese segmenter with the web demo someday…):
Jacy won’t parse it without a question particle:
With the question particle (or complementizer, really) か ka on the embedded clause, it gets 122 results:
I’m not sure if all 122 are legit, though.
Other complementizers, like と to (or both, か と ka to) also work, but I think と to is infelicitous in this case. I think it’s ungrammatical to omit a complementizer on the embedded question, but perhaps it’s ok in casual speech.